Autonomous vehicles: Behavioral considerations

Once you have determined the key policies that are of interest to your agency, the next step is to think through the potential behavioral implications of these policies on travel, activity, and land use. In particular, on how the existence of AV may impact these behaviors. The essential change to the transport system that CAVs bring is their potential to make auto travel more convenient on a number of levels. As one no longer has to execute the driving task, this reduces driver fatigue, increases safety, and allows for the traveler to engage in other activities (multitasking). Further, as the car can travel without a human driver backup, the car can drop travelers directly at the destination, be sent on errands, carry those who cannot themselves drive (children, disabled, elderly), and relocate itself to coordinate travel for multiple people. In the absence of opposing forces such as congestion, pricing, infrastructure prioritization, or major land use restructuring, this increasing ease of auto travel will likely lead to more consumption (i.e. more auto travel).

To begin to understand the magnitude of the increase in travel, and the potential for policy to influence this increase, it is necessary to dig down into the specific behavioral shifts from which this increased travel occurs. In discussing such behavioral shifts, it is useful to separate the impacts on short-term, medium-term, and long-term behavioral decisions. The “length” of the decision reflects how easy it is for a person to make adjustments, and therefore how quickly such behavioral changes will be realized in the system after CAVs are introduced. For example, leaving 30 minutes earlier is a relatively easy adjustment, so classified as a short-term decision that is expected to happen relatively quickly after the introduction of CAVs (months); moving residential location is hard and happens less frequently, so a long-term decision that will take some time (years, decades) for adjustments to be realized after CAVs are introduced. Of course, the aggregate scale of any of these adjustments is a function of the diffusion rate of CAVs in the system.

# Short-term behavioral decisions

The short-term behaviors are the daily decisions of what activities to engage in, and how and when to access these activities. In a classic modeling framework, this includes selection and location of discretionary activities (i.e., non-work) and the timing, mode, and route by which all activities (including mandatory activities) are accessed. All of these aspects may be influenced by the introduction of AVs. The increased ease of travel by auto may lead to more trips generated (i.e., activity/trip generation), for example by those who currently are not able to drive (due to age restrictions or disabilities); by lowering the impedance of more difficult trips such as longer trips, evening trips, activities leading to driving impairment (e.g., alcohol); by sending the AV on errands; or by replacing online ordering/delivery with AV pickup. The introduction of CAVs will also impact mode choices, shifting more travel to auto from other modes. The alternatives in the CAV future to which these shifts will occur will include options for self-owned CAVs as well as fleets of shared CAVs that can be used either solo or being pooled with other travelers. The CAVs can either be used for the entire trip or to link (either access or egress) to transit alternatives. The destinations of activities are also likely to change, with the increased ease of auto travel leading to longer trips (i.e., farther activities). Finally, the timing of trips may adjust as the sensitivity, for example, to peak period congestion may be lessened. The route can also be impacted depending on restrictions of roadway segments. More generally, there may be broader changes in trip-chaining and tours due to the ability of the CAV to shuttle multiple people among destinations, to relocate independently, and to drive people who used to have to be driven. For example, members of a household may coordinate complex, multi-person activity schedules via the use of a single household CAV.

# Medium-term behavioral decisions

The critical medium-term behavioral decisions in the CAV space relate to decisions regarding the household mobility bundle. This includes both the number and types of vehicles to own (both conventional and CAV vehicles as well as other vehicle options such as bikes) and also passes or memberships for transit services or shared mobility systems. The introduction of CAVs may lead to households purchasing CAVs, perhaps shedding a vehicle if they can achieve the same level of mobility by coordinating with fewer vehicles. Alternatively, it may lead to households forgoing personal auto ownership and electing for use of a shared CAV fleet.

Other medium-term behavioral decisions that may be impacted by CAVs are the destination and timing of frequent activities (such as child care and classes) as well as the decision to telework or not. These decisions are impacted in the same manner as described for short-term decisions, but they tend to be made on a monthly or longer basis rather than on a daily basis.

# Long-term behavioral decisions

The long-term decisions relate to the land use feedback that is driven by the changing level of service of the transport system and therefore changes in accessibility. The key household decisions here are residential location choice and workplace location choice. The increasing ease of auto travel makes it less onerous to live farther from one’s activity locations, and therefore people may do so.

The bundle of travel-related choices, be they short-, medium-, or long-term, made by an individual or household represent a modality style choice, which is a lifestyle built around the use of a particular travel mode or set of travel modes. For example, there are auto-oriented individuals that for the most part travel by car and their choices (destination, mode, vehicles, housing, etc.) reflect this orientation. Conversely, there are households with more of a multi-modal orientation, and make decisions from residential choices, to vehicle to mode choice based on this orientation. The impact of CAVs on such higher-order modality styles (e.g., either towards or away from higher auto-orientations) will largely guide the resulting demand and performance of the transportation system. Operational models today do not explicitly represent such modality styles by bundling choices and making clear segments of the market, but they do reflect the key modality-style decisions in terms of auto ownership and housing location choices.

# Reflecting behavioral impacts in travel demand models

In some ways, the potential behavioral adjustments that are described above are reflected in our current paradigms of travel demand modeling via sensitivities to time and cost of alternatives and implicit reflection of convenience and modal preferences via modal constants. Of course, the devil is in the details in terms of defining and adding new alternatives and determining the adjustments necessary to the parameters on time, cost, and modal constants. Perhaps the more difficult behavioral adjustments to capture are those that reflect relatively new behaviors that are made possible by the introduction of CAVs, in particular enabled by the CAVs ability to travel without a human backup. For example, the new possibilities of coordination in terms of timing and trip-chaining (and dead-heading) within a household, group, or community sharing privately-owned CAVs; the use of the vehicle for populations that currently cannot drive; and the ability of the car to park itself and run errands.

The obvious complication of modifying our demand models to predict travel behavior with CAVs is the simple reality that CAVs do not currently exist. However, there is a rich travel behavior literature that is being leveraged on related topics such as pricing, multitasking, parking, travel budgets, modal attitudes, habits, etc. With this knowledge base, there are a wide range of adjustments that can be made to the models to reflect CAV impacts. These include making adjustments to model parameters (such as increasing trip generation rates or reducing travel time sensitivities in mode, destination, and residential location models), reducing alternative availability limitations (e.g., allowing independent private auto use for people without licenses), adding new alternatives (such as CAV in a mode choice or auto ownership model), or developing new behavioral models (such as automated parking behavior, errand running by CAV, intra-household coordination of household CAV, or privately owned vs shared CAV use). Further, current modeling efforts on TNCs (Uber, Lyft, etc.) and the infusion of these new mobility alternatives into the models help provide a framework for introducing CAV elements into travel demand models. In addition, new behavioral experiments are being developed to particularly target CAV behaviors. These include stated preference and willingness to pay studies asking about hypothetical CAV future scenarios, field tests using current day analogies (e.g., TNCs and chauffeurs), and gaming and virtual reality. As more of behavioral knowledge is acquired and compiled, more thoughtful adjustments can able to be made to the travel demand models. Some variables that are critical to forecasting the potential outcome are at this point very difficult to predict, such as the relative market shares of privately-owned CAVs and shared CAVs (which is also particularly contentious). For parameters such as this, it perhaps makes more sense to treat it as an exogenous, scenario level variable rather than endogenously predicting it. The path forward requires a careful balance of adjusting what we can now based on current knowledge, and updating our approaches as new behavioral insights are gained and as new models are developed.

While here we presented a high-level discussion of model development, specific and tactical details on modifying travel demand models can be found in modeling frameworks.

# References

Content Charrette: Autonomous Vehicles

This site uses cookies to learn which topics interest our readers.